Deep Generalized Learning Model for PET Image Reconstruction.

Low-count positron emission tomography (PET) imaging is challenging because of the ill-posedness of this inverse problem. Previous studies have demonstrated that deep learning (DL) holds promise for achieving improved low-count PET image quality. However, almost all data-driven DL methods suffer from fine structure degradation and blurring effects after denoising. Incorporating DL into the traditional iterative optimization model can effectively improve its image quality and recover fine structures, but little research has considered the full relaxation of the model, resulting in the performance of this hybrid model not being sufficiently exploited. In this paper, we propose a learning framework that deeply integrates DL and an alternating direction of multipliers method (ADMM)-based iterative optimization model. The innovative feature of this method is that we break the inherent forms of the fidelity operators and use neural networks to process them. The regularization term is deeply generalized. The proposed method is evaluated on simulated data and real data. Both the qualitative and quantitative results show that our proposed neural network method can outperform partial operator expansion-based neural network methods, neural network denoising methods and traditional methods.

Improved [68Ga]Ga-PSMA-11 image qualities reconstructed by total variation regularized expectation maximization on total-body PET/CT.

Background: Total variation regularized expectation maximization (TVREM) reconstruction algorithm on the image quality of gallium (68GA) prostate-specific membrane antigen-11 ([68Ga]Ga-PSMA-11) total-body positron emission tomography/computed tomography (PET/CT). Methods: Images of a phantom with small hot sphere inserts and the total-body PET/CT scans of 51 prostate cancer patients undergoing [68Ga]Ga-PSMA-11 were reconstructed using TVREM with 5 different penalization factors between 0.09 and 0.45 and for 20-, 40-, 60-, 120-, and 300-second acquisition, respectively. As a comparison, the same data were also reconstructed using the ordered subset expectation maximization (OSEM) with 3 iterations, 20 subsets, and 300 second acquisition. The contrast recovery coefficients (CRC) and background variability (BV) of the phantom, the tumor-to-background ratios (TBR), the contrast recovery (CR) ratio, the image noise of the liver, and maximum standard uptake value (SUVmax) of the lesions were calculated to evaluate the image quality. The clinical performance of the images was evaluated by 2 radiologists with a 5-point scale (1-poor, 5-excellent). Results: The TVREM reconstructions groups fwith 120 second acquisition and the penalization of 0.27 to 0.45 showed the best performance in terms of CR, TBR, image noise, and the gain of SUVmax compared to that obtained in the OSEM 300 second group. Even the image noise of the TVREM 120 second group with a penalization factor of 0.27 and 0.36 was comparable to the OSEM 300 second group; the lesions' SUVmax increased by 28% whereas the image noise decreased by 5% and 14%, respectively. The TVREM 120 second group with a penalization factor of 0.36 (5.00±0.00) had the highest qualitative score that equaled OSEM and TVREM for the 300 second (P>0.05) group. Conclusions: Our study has shown the potential of the TVREM reconstruction algorithm with optimized penalization factors to achieve comparable [68Ga]Ga-PSMA-11 total-body PET/CT image quality with a shorter acquisition time, compared with the conventional OSEM reconstruction algorithm.

Motion correction and its impact on quantification in dynamic total-body 18F-fluorodeoxyglucose PET.

BACKGROUND: The total-body positron emission tomography (PET) scanner provides an unprecedented opportunity to scan the whole body simultaneously, thanks to its long axial field of view and ultrahigh temporal resolution. To fully utilize this potential in clinical settings, a dynamic scan would be necessary to obtain the desired kinetic information from scan data. However, in a long dynamic acquisition, patient movement can degrade image quality and quantification accuracy. METHODS: In this work, we demonstrated a motion correction framework and its importance in dynamic total-body FDG PET imaging. Dynamic FDG scans from 12 subjects acquired on a uEXPLORER PET/CT were included. In these subjects, 7 are healthy subjects and 5 are those with tumors in the thorax and abdomen. All scans were contaminated by motion to some degree, and for each the list-mode data were reconstructed into 1-min frames. The dynamic frames were aligned to a reference position by sequentially registering each frame to its previous neighboring frame. We parametrized the motion fields in-between frames as diffeomorphism, which can map the shape change of the object smoothly and continuously in time and space. Diffeomorphic representations of motion fields were derived by registering neighboring frames using large deformation diffeomorphic metric matching. When all pairwise registrations were completed, the motion field at each frame was obtained by concatenating the successive motion fields and transforming that frame into the reference position. The proposed correction method was labeled SyN-seq. The method that was performed similarly, but aligned each frame to a designated middle frame, was labeled as SyN-mid. Instead of SyN, the method that performed the sequential affine registration was labeled as Aff-seq. The original uncorrected images were labeled as NMC. Qualitative and quantitative analyses were performed to compare the performance of the proposed method with that of other correction methods and uncorrected images. RESULTS: The results indicated that visual improvement was achieved after correction of the SUV images for the motion present period, especially in the brain and abdomen. For subjects with tumors, the average improvement in tumor SUVmean was 5.35 ± 4.92% (P = 0.047), with a maximum improvement of 12.89%. An overall quality improvement in quantitative Ki images was also observed after correction; however, such improvement was less obvious in K1 images. Sampled time-activity curves in the cerebral and kidney cortex were less affected by the motion after applying the proposed correction. Mutual information and dice coefficient relative to the reference also demonstrated that SyN-seq improved the alignment between frames over non-corrected images (P = 0.003 and P = 0.011). Moreover, the proposed correction successfully reduced the inter-subject variability in Ki quantifications (11.8% lower in sampled organs). Subjective assessment by experienced radiologists demonstrated consistent results for both SUV images and Ki images. CONCLUSION: To conclude, motion correction is important for image quality in dynamic total-body PET imaging. We demonstrated a correction framework that can effectively reduce the effect of random body movements on dynamic images and their associated quantification. The proposed correction framework can potentially benefit applications that require total-body assessment, such as imaging the brain-gut axis and systemic diseases.

Evaluation of pediatric malignancies using total-body PET/CT with half-dose [18F]-FDG.

PURPOSE: To explore the impact of a true half dose of [18F]-FDG on image quality in pediatric oncological patients undergoing total-body PET/CT and investigate short acquisition times with half-dose injected activity. METHODS: One hundred pediatric oncological patients who underwent total-body PET/CT using the uEXPLORER scanner after receiving a true half dose of [18F]-FDG (1.85 MBq/kg) were retrospectively enrolled. The PET images were first reconstructed using complete 600-s data and then split into 300-s, 180-s, 60-s, 40-s, and 20-s duration groups (G600 to G20). The subjective analysis was performed using 5-point Likert scales. Objective quantitative metrics included the maximum standard uptake value (SUVmax), SUVmean, standard deviation (SD), signal-to-noise ratio (SNR), and SNRnorm of the background. The variabilities in lesion SUVmean, SUVmax, and tumor-to-background ratio (TBR) were also calculated. RESULTS: The overall image quality scores in the G600, G300, G180, and G60 groups were 4.9 ± 0.2, 4.9 ± 0.3, 4.4 ± 0.5, and 3.5 ± 0.5 points, respectively. All the lesions identified in the half-dose images were localized in the G60 images, while 56% of the lesions could be clearly identified in the G20 images. With reduced acquisition time, the SUVmax and SD of the backgrounds were gradually increased, while the TBR values showed no statistically significant differences among the groups (all p > 0.1). Using the half-dose images as a reference, the variability in the lesion SUVmax gradually increased from the G180 to G20 images, while the lesion SUVmean remained stable across all age groups. SNRnorm was highly negatively correlated with age. CONCLUSION: Total-body PET/CT with a half dose of [18F]-FDG (1.85 MBq/kg, estimated whole-body effective dose: 1.76-2.57 mSv) achieved good performance in pediatric patients, with sufficient image quality and good lesion conspicuity. Sufficient image quality and lesion conspicuity could be maintained at a fast scanning time of 60 s with half-dose activity.

[Incorporation of TV regularization in deconvolution for partial volume correction in PET imaging].

OBJECTIVE: We propose a method using total variation (TV) regularization in deconvolution for partial volume correction in PET imaging. In the degraded image model, we used TV regularization procedure in Van Cittert (VC) and Richardson-Lucy (RL) deconvolution algorithms. These methods were tested in simulated NCAT images and images of NEMA NU4-2008 IQ phantom and tumor-bearing mouse scanned by Simens Invoen microPET. The simulated experiment and tumor-bearing mouse experiment showed that the algorithms using TV regularization provided superior qualitative and quantitative appearance compared with traditional VC and RL algorithms. When the mean intensity of the tumor increased by (10±1.8)%, the SD increase percentage was decreased from 49.98% to 14.26% and from 42.76% to 4.70%, suggesting the efficiency of the proposed algorithms for reducing PVEs in PET.

[Edge-detecting operator-based selection of Huber regularization threshold for low-dose computed tomography imaging].

OBJECTIVE: To compare two methods for threshold selection in Huber regularization for low-dose computed tomography imaging. METHODS: Huber regularization-based iterative reconstruction (IR) approach was adopted for low-dose CT image reconstruction and the threshold of Huber regularization was selected based on global versus local edge-detecting operators. RESULTS: The experimental results on the simulation data demonstrated that both of the two threshold selection methods in Huber regularization could yield remarkable gains in terms of noise suppression and artifact removal. CONCLUSION: Both of the two methods for threshold selection in Huber regularization can yield high-quality images in low-dose CT image iterative reconstruction.

Threshold choices of Huber regularization using global- and local-edge-detecting operators for X-ray computed tomographic reconstruction.

Statistical iterative reconstruction (SIR) approaches have shown great potential in x-ray computed tomographic (CT) reconstruction in the case of low-dose protocol. For yielding high quality image, an edge-preserving regularization should be incorporated into the objective function of SIR approaches. A typical example is the Huber regularization with an edge-preserving non-quadratic potential function which increases less rapidly than the quadratic potential function for sufficiently large arguments. However, a major drawback of the Huber regularization is the determining the threshold, which precludes its extensive applications. In this paper, we investigate both global- and local- edge-detecting operators for threshold choices of Huber regularization and apply them to SIR CT image reconstruction with low-dose scan protocol. Experiments were performed on XCAT phantom by using a CT simulator to obtain the low-dose projection data.